Fluctuations in State-Level Monthly Unemployment Rates: A Descriptive Analysis of the Effect of Geographic Location and U.S. Unemployment Rates

This study examines fluctuations in state-level monthly unemployment rates from January 1976 to December 2016, a period of 492 months. Using data from the Bureau of Labor Statistics (BLS) Local Area Unemployment Statistics series, patterns in the monthly seasonally adjusted unemployment rates of the 50 states and the District of Columbia are examined. The study focuses on two unemployment-related issues. First, the relationship between a state’s unemployment rate and the U.S. unemployment rate is examined. Second, we explore the extent to which a state’s geographic location, using Census regions and Census divisions, affects its monthly unemployment rate.

suggests that both the hysteresis and structuralist models of state unemployment can be useful in explaining state unemployment, but that the time period under consideration accounts heavily into which view is more accurate. He concludes that over the last 30 years most state unemployment rates were characterized by stationarity fluctuations around a shifting trend. A paper that presents an interesting comparison of US state unemployment rates and European Union unemployment rates is Romero-Avila and Usabiaga (2009). The authors employ panel stationarity tests to examine the main unemployment paradigms, looking specifically at data from the US states and the EU over roughly the last 30 years. The authors' findings also suggest that US state unemployment is characterized by stationarity, whereas EU unemployment rates exhibit more closely the hysteresis paradigm, and they conclude that adverse macroeconomic shocks, such as interest rates and oil prices, cause permanently higher unemployment rates in the EU than in the states. Owyang, Piger, and Wall (2005) use the state coincident index, like Ewing and Thompson (2012), to present evidence regarding the timing and characteristics of state-level business cycles. They use monthly data from January 1979 through June 2002 and determine that there are large differences across states in both the contraction phases and expansions phases of the cycle. They further find that the differences in recession growth rates among states tend to depend on differences in the employment-mix characteristics of the states, but that differences in expansion growth rates tend to depend on differences in demographic characteristics across states. Their findings also indicate that national recessions "are less reflective of middle America and more indicative of the East and Far www.scholink.org/ojs/index.php/ibes International Business & Economics Studies Vol. 2, No. 2, 2020 37 Published by SCHOLINK INC.
Previous research has also studied state unemployment rates relative to national economic conditions based on the industrial characteristics of the states. Partridge and Rickman (1997) explore the differences in state unemployment rates as a function of national economic variables and state equilibrium and disequilibrium variables. The authors define state equilibrium variables as those that affect state unemployment differentials after state growth rates are equalized (i.e., normalized to a national trend), thus reducing the flow of resources from state to state. Disequilibrium variables cause state unemployment rates to diverge because of state industry mix factors and competitive shift between states. The authors conclude that the relative importance of equilibrium versus disequilibrium differs dramatically across states and regions, and knowledge of these differences could aid policy makers Hyclak and Lynch (1980) use a version of Okun's Law to measure the impact of the U.S. business cycle on state unemployment rates. Their empirical results show a high degree of variability and sensitivity of state unemployment rates to the national output gap. In addition, they find that the industrial mix of states has a significant effect on unemployment sensitivity. States with large manufacturing and tourism sectors have higher unemployment sensitivity than states that have large cities, or significant agricultural sectors. Izraeli and Murphy (2003) examine the impact of industrial diversity on state unemployment rates and per-capita income. The authors argue that industrial diversification can serve as a type of unemployment insurance during downturns in the business cycle.
Using panel data for seventeen states, they find a strong link between industrial diversity and reduced unemployment. Walden (2012) examines industrial composition and its impact on state unemployment.
He concludes that during the Great Recession industrial composition (manufacturing in particular), falling housing prices, and household in-migration contributed to state unemployment. Walden further finds that the states most severely impacted by these three economic characteristics were geographically clustered in three regions, the Far West, the Southeast, and the Midwest. Beyers (2013) also examines the industrial characteristics of states and how these conditions contributed to state unemployment during the Great Recession. He finds that states with higher concentrations of manufacturing tended to suffer higher rates of unemployment during the recession.
Nistor (2009) also attempts to identify state-level factors that affect a state unemployment rates, using state-level annual unemployment data for 1990 and 2000, and finds that human capital investment in a state negatively affects its unemployment rate. Vedder and Gallaway (1996)  state unemployment as a function of the labor market freedom in individual states. He finds that states with greater labor market freedom tend to have lower unemployment rates. According to Cebula, in order to increase labor market freedom (and therefore lower unemployment), states should reconsider policies regarding minimum wages, government employment, and right-to-work laws.
This paper differs from the previous literature in that it focuses only on the relationship between state-level unemployment rates and national unemployment rate, and on the effect that a state's geographic location has on its unemployment rate. It also determines how the effect of the U.S. unemployment rate on a state unemployment rates differs across states.

The Data and Descriptive Statistics
The state-level data used in this study are obtained from the U.S. Bureau of Labor Statistics, "Local    Central Census division and is lowest (i.e., the most negative) for states in the West North Central Census division.  A set of graphs that plots the difference between each state's unemployment rate and the U.S. unemployment rate for each month is shown in Figure 1.        In the following section, we examine this issue in more detail by estimating a series of econometric models, where geographic location (Census region, Census division, or state) are explanatory variables.
Based on the above analysis, it's our expectation that geographic location will be a strong determinant of state-level monthly unemployment rates. We expect to find that states located outside the Midwest Census region and outside the West North Central Census division to have relatively high monthly unemployment rates, ceteris paribus.

Econometric Models
Two different sets of regression equations are estimated in this paper. In the first set of models, all 50 states and the District of Columbia are included as observations in a single regression equation. To ascertain whether there are differences in monthly unemployment rates across geographic space, dummy variables related to the location of a state are included as explanatory variables in the models.
The general form of these regressions is: (1) Rate i, t = a + b US t + D i where Rate i,t is the seasonally adjusted unemployment rate for state i in month t.
US t is the U.S. seasonally adjusted unemployment rate in month t. D i is geographic dummy variable to denote the location of state i. omitted region/division/state has a mean monthly unemployment rate that's lower than that of the other regions/divisions/states.
In the second set of models, a separate regression is run for each state and the District of Columbia, which yields 51 separate regression equations. The general form of these regressions is: (2) Rate t = a + b US t where Rate t is the seasonally adjusted unemployment rate for the state in month t.
US t is the U.S. seasonally adjusted unemployment rate in month t.

Discussion of Regression Results
The results of equation (1)   Note. The absolute values of the t-statistics are in parenthesis.
In the State Dummies model all state-dummy coefficients are positive and are statistically significant at the .01 level. Nebraska, the state with the lowest mean unemployment rate, is the omitted state.
The results of equation (2) are reported in  .249 Note. The absolute values of the t-statistics are in parenthesis.
The coefficient on the US Rate is statistically significant at the .01 level in all 51 equations.

Predicted Unemployment Rates
In Table 7, predicted unemployment rates for each state are reported for six different U.S. unemployment rates, using the regression results from Table 6. The U.S. unemployment rates that are utilized are 4.0%, 5.0%, 6.0%, 7.0%, 8.0% and 9.0%. For a historical context, the U.S. unemployment rate was 9.0 percent or higher 49 times during the 492-month period considered in this study, and it was 4.0 percent or lower 11 times. As such, these two rates are relatively rare.
The predictions indicate a wide range of state unemployment rates for a given U.S. unemployment rate.
The predictions also indicate that the range between the unemployment rate of the state with the highest rate and that with the lowest rate is likely to increase as the U.S. unemployment rate increases. When the U.S. unemployment rate is an unusually low 4.0 percent, the range in predicted state unemployment rates is 4.27 percentage points, but when the U.S. unemployment rate is an unusually high 9.0 percent, the predicted range is a much larger 7.81 percentage points.
The predictions also indicate that some states are likely to always have an unemployment rate that's lower than the U.S. rate (Colorado, Connecticut, Delaware, Nebraska, North Dakota, South Dakota, and Utah, for example), while others are likely to always have an unemployment rate that's above the U.S. rate (Alaska, California, District of Columbia, Illinois, Mississippi, and West Virginia, for example).